Tracking disease progression by searching paths in a temporal network of biological processes

Abstract

Metabolic disorders such as obesity and diabetes are diseases which develop gradually over time through the perturbations of biological processes. These perturbed biological processes usually work in an interdependent way. Systematic experiments tracking disease progression at gene level are usually conducted through a temporal microarray data. There is a need for developing methods to analyze such highly complex data to capture disease progression at the molecular level. In the present study, we have considered temporal microarray data from an experiment conducted to study development of obesity and diabetes in mice. We first constructed a network between biological processes through common genes. We analyzed the data to obtain perturbed biological processes at each time point. Finally, we used the biological process network to find links between these perturbed biological processes. This enabled us to identify paths linking initial perturbed processes with final perturbed processes which capture disease progression. Using different datasets and statistical tests, we established that these paths are highly precise to the dataset from which these are obtained. We also established that the connecting genes present in these paths might contain some biological information and thus can be used for further mechanistic studies. The methods developed in our study are also applicable to a broad array of temporal data.

Extracting proteins involved in disease progression using temporally connected networks

“…Briefly, out of total ~ 29,411 genes measured, data was processed to filter out genes not perturbed 2 fold even in one time point resulting in a matrix of 19,303 genes across 10 time points. Further details of the processing are given in [ 9 ].…”

“…We recently published a similar work [ 9 ] tracking disease progression where the biological processes perturbed at each time point were found and connected using a network of connected biological processes The work analyses progression of data by finding processes perturbed temporally and connecting them resulting in paths, giving proteins of processes of the paths as gene/protein targets. A more precise assessment of importance could be firstly quantitatively defining progression and then quantifying the effect of removal of the protein from the network on progression, which was missing in the previous work [ 9 ]. Here, we address this limitation by making a novel algorithm giving a more detailed quantitative assessment of the importance of the protein in question.…”

VD-Analysis: A Dynamic Network Framework for Analyzing Disease Progressions

“…Aiming at understanding the pathogenesis of heart disease based on time course RNA-seq dataset, Ma et al proposed the inference of multiple differential modules (iMDM) algorithm to identify gene modules across multiple differential coexpression networks [8]. Based on temporal microarray data of gene expression, Rajat et al obtained dysregulated gene modules at 10 time points, which constitute paths of the disease evolution [9]. Srihari et al identified cancer-related protein complexes via analyzing expression level differences for normal and cancer conditions [10].…”

Restless reachability problems in temporal graphs

“…Graphs are used to study many real-world problems such as information propagation in social networks [6,23], spreading of epidemics [3,30], protein interactions [1,31], activity in brain networks [12,15], and more. Over the years the graph models are enriched with information of node and edge attributes and edge timestamps, giving rise to attributed graphs [29,35] and temporal graphs [22,28], which are used to capture complex phenomena and network dynamics.…”